I build and maintain Rails applications for US and Mexican companies. Eight years of senior judgment, in your time zone, now with Claude Code at the keyboard.
I take over codebases other developers left behind, stabilize them, and bring them up to current Rails and Ruby without breaking what already works.
For apps on an old Rails version, teams whose original developer is gone, or anyone afraid to touch deploys.
A written spec you approve, AI-assisted implementation with a senior engineer reviewing every line, and a tested, deployed app in a fraction of the usual timeline.
For validating a product, replacing a no-code tool, or an MVP that won’t need a rewrite.
Live updates, inline editing and smooth navigation with Turbo and Stimulus. One codebase, one deploy, and a UI your users will assume was built with React.
For apps that feel dated, or teams weighing a SPA rewrite who want a cheaper path.
Marketing sites and documentation that load instantly, score high on Lighthouse, and cost almost nothing to host. Editable through Markdown or a headless CMS.
For a fast, SEO-friendly site for a product, a business, or a campaign.
I built an open-source Rails skill for Claude Code that encodes the conventions, guardrails and patterns I’ve refined over eight years, so the AI’s output already looks like a senior Rails engineer wrote it.
For you that means faster delivery, fewer surprises, and a codebase the next developer can read without a translation layer.
This is the flow I follow to build and maintain Ruby on Rails software with Claude Code. It’s Anthropic’s AI-native SDLC, adapted to Rails: the AI runs the loop, I review every artifact it produces, and nothing reaches production without my sign-off.
01
Your idea becomes a written intent file. We brainstorm it with Claude until it’s a structured proto-spec, and it goes into version control from day one.
02
Requirements and design collapse into one session. Claude drafts the spec guided by my Rails skill, so conventions, security and data rules are applied while it’s written.
03
Every task starts in plan mode. I approve the plan before a line of code exists, and a CLAUDE.md with your project’s knowledge keeps the agent inside your codebase’s rules.
04
The session verifies its own work with RSpec, the build and screenshots before I ever see it. What lands in my review already passes, so I spend my time on judgment.
05
Claude reviews incoming pull requests against the project’s policies and answers review comments on its own. Hooks enforce the gate: it can work up to production, never past it.
06
Errors and monitoring alerts trigger a diagnosis and a proposed fix. Findings re-enter as a new intent, and the loop starts again.
The loop keeps running. Human judgment stays above it.
Based on Anthropic’s AI-native SDLC playbook.
Your standup is my standup. Central Time overlaps the full US working day, so questions get answered the same morning, not the next.
Senior rates from Mexico are a fraction of US contractor rates, with no agency in between.
Technical conversations happen in fluent English or native Spanish, whichever your team prefers.
Eight years of Rails. The speed of the AI era.
Five fields, and a reply within one business day. Prefer email? Write to francisco@tolesco.com.
Message sent. Talk soon.
I’ll reply within one business day with a link to book a call.
